Calibrating Nonlinear Mobile Sensors

In-field calibration of sensor devices is known to be a challenging problem because there is often no access to a controlled signal field and/or a pre-calibrated device to provide the ground truth. Nonlinear characteristics of sensor devices make the calibration problem even harder. In this paper, we describe two blind calibration schemes for nonlinear mobile sensor nodes: nullspace based calibration (NBC) and moments based calibration (MBC). Simulation results are included to demonstrate the effectiveness of the proposed schemes. MBC scheme is also used to calibrate light sensors on MICA2 motes in a light field generated by a light bulb. Results show that significant error reduction can be achieved when nonlinearity is considered.